Re: Information components (DTD fragments)

Nobody can possibly state that this stuff isn't important.  I would
reinforce Jon's point; it is *so* important it needs to go 
either in to XML-lang or (more likely) into an ancillary spec to
be entitled "Metadata and Declarations for XML".  There are a lot
of things to think about.  To start with, we are *not* limiting
our scope to the kind of things DTD's talk about.  DTD's are 
simultaneously much too limiting (I often don't care that much
about order and nesting of well-defined info elements) and woefully
incomplete, saying nothing about semantics, data typing, behavior,
or availability.  So when we say "declarations" we need to talk
about a lot more than content models and ATTLISTs.

Then there's the name-space problem.  I envision a future in which
you can copy in Oenology declarations (vintages, varieties, chateaux)
from one place and Foreign Exchange declarations from another, and
put together a whizzy wine-merchant-on-the-net document package.  But
you have a world-class hairy namespace problem.  I don't *think* it's
deeply architectural, it just requires some clever plumbing, but 
it's real.

Then there's the big issue - behavior - which means having a general
way to find & use "methods" associated with elements.  Peter Murray-Rust's 
technique - element "foo" is handled by "foo.class" - is appealingly simple, 
but I think people are going in general to want a couple more layers of
indirection, and probably to use IDL rather than Java.

I think we can guarantee that this is going to come up in an official
way, and probably cause an unprecedented level of involvement from
the big browzer boyz, W3C heavies, and so on.  But we gotta
get the XML-link basics cleaned up and some basic stylesheet linkage
before we can really wade into it. -Tim

Received on Sunday, 4 May 1997 23:54:57 UTC